Biochemistry 1993, 32, 2987-2998 2987
Binding of 4’,6-Diamidino-2-phenylindole(DAPI) to AT Regions of DNA: Evidence for an Allosteric Conformational Change
Svante Eriksson, Seog K. Kim, Mikael Kubista,’ and Bengt Nordtn Department of Physical Chemistry, Chalmers University of Technology, S-412 96 Gbteborg, Sweden
Received July 9, 1992; Revised Manuscript Received November 6, 1992
ABSTRACT: The interaction of 4/,6-diamidino-2-phenylindole(DAPI) with several double-helical poly- and oligonucleotides has been studied in solution using optical spectroscopic techniques: flow linear dichroism (LD), induced circular dichroism (CD), and fluorescence spectroscopy. In AT-rich sequences, where DAPI is preferentially bound, LD indicates that the molecule is edgewise inserted into the minor groove at an angle of approximately 45O to the helix axis. This binding geometry is found for very low as well as quite high binding ratios. The concluded geometry is in agreement with that of the DAPI complex in a crystal with the Drew-Dickerson dodecamer, and the DAPI complex with this dodecamer in solution is verified to have an ICD spectrum similar to that of the complex with [poly(dA-dT)]z at low binding ratios. The observation of two types of CD spectra characteristic for the binding of DAPI to DNA, and also for the interaction with [poly(dA4T)]2,demonstrates that the first binding mode, despite its low apparent abundance (a few percent), is not due to a specific DNA site. The effect may be explained in terms of an allosteric binding such that when DAPI molecules bind contiguously to the AT sequence the conformation of the latter is changed. The new conformation, which according to LD appears to be stiffer than normal B-form DNA, is responsible for the second type of induced CD spectrum in the DAPI chromophore. Although the spectroscopic results indicate a change of DNA conformation, consistent with an allosteric binding model, they do not explicitly require any cooperativity, but accidental neighbors could also explain the data.
The interaction of DAPI’ (4’,6-diamidino-2-phenylindole) of the geometry of this complex between DAPI and DNA is with DNA has been the subject of numerous studies since this provided from the crystal structure of DAPI bound to the drug was first synthesized (Dann et al., 1971). The original dodecamer, [d(CGCGAATTCGCG)]2, where it is found in intent was to produce a diamidino compound, analogous to the minor groove of the central AATT sequence (Larsen et berenil and stilbamidine, to be used as a trypanocide agent. al., 1989). However, the special spectral properties of DAPI have caused Most studieshave concerned this high-affinity binding mode, it to be used more as a DNA probe. It has been exploited, but heterogeneity in the binding has been evidenced at higher for example, in electrophoresis (Kapuscinski & Yanagi, 1979; binding ratios both with DNA and with some polynucleotides Schwartz & Koval, 1989), for cytofluorometry (Brown & (Bierzynski et al., 1978; Kapuscinski & Szer, 1979; Kubista Hitchcook, 1989; Chi et al., 1990; Hajduk, 1976; Lee et al., et al., 1987; Manzini et al., 1983, 1985; Norddn et al., 1990; 1984; Russel et al., 1975; Takato & Hirano, 1990; Tijssen et Wilson et al., 1990). Several explanations have been put al., 1982; Williamson & Fennel, 1979) and for staining forward, including an allostericchange of the DNA structure chromosomes (Bella & Goshlvez, 1991; Bernheim & Miglie- (Norddn et al., 1990; Wilson et al., 1990). Wilson and co- rina, 1989; Lin et al., 1977). workers (1989) have shown that DAPI can bind to GC DAPI forms highly fluorescent complexes with double- sequences too, possibly by an intercalative binding mode. stranded DNA (Cavatorta et al., 1985; Kapuscinski & The structural interpretation of spectroscopicdata of DAPI- Skoczylas, 1978; Williamson & Fennell, 1975, 1979) with a DNA complexes requires knowledge of the basic electronic binding preference for AT-rich regions (Williamson & Fennel, properties of the ligand chromophore. The broad near-UV 1975). Footprinting experiments show that a sequence of absorption of DAPI originates from two electronic transitions; 3-4 contiguous AT base pairs is covered when DAPI is bound one is dominating the long wavelength region and is polarized (Jeppesen & Nielsen, 1989; Portugal & Waring, 1988). along the molecular long axis of DAPI, and one at slightly Different techniques such as fluorescence (Htrd et al., 1990; shorter wavelength is polarized some 15O to the former Linet al., 1977),linear dichroism (Kubista et al., 1987),NMR (Kubista et al. 1989). In aqueous solution their absorption (Wilson et al., 1990), and viscometry (Manzini et al., 1985) maxima are centered around 363 and 334 nm, respectively. suggest that DAPI is located in the minor groove of the DNA The two transitions involve both the indole and the phenol helix in this high-affinity binding mode. A detailed description chromophores (Albinsson et al., 1991). The photophysics of DAPI is complex, probably involving both ground-state and * Author to whom correspondence should be addressed. excited-state heterogeneities (Barcellona & Gratton, 1990, Abbreviations: A, adenine; C, cytidine;CD, circular dichroism;DAF’I, 1991). The large increase in the fluorescence quantum yield 4’,6-diamidino-2-phenylindole;DNA, deoxyribonucleic acid; EDTA, of DAPI upon binding to AT regions has been attributed to ethylenediaminetetraaceticacid; G, guanine; LD, linear dichroism; LD’, hindrance of an intramolecular proton transfer in the excited reduced linear dichroism; [poly(dA-dT)]2, duplex of alternating poly- (dA-dT); [poly(dGaC)]z, duplex of alternating poly(dG4C); T, state, which in solution gives rise to rapid quenching of the thymine; Tris, tris(hydroxymethy1)aminomethane: A, wavelength. fluorescence (Szabo et al., 1986). 0006-2960/93/0432-2987$04.00/0 0 1993 American Chemical Society 2988 Biochemistry, Vol. 32, No. 12, 1993 Eriksson et al. Despite extensive studies of DAPI's interaction with DNA, Linear dichroism (LD) is defined as a number of questions have remained unanswered concerning the geometrical prerequisites for its binding, i.e., the sequence WX) = A,,(X)-A,@) (1) dependence, and how the geometries differ between various where All(X) and AI(X)are the absorption spectra measured modes of binding. The objectives of this study are to with the light plane polarized parallel and perpendicular, characterize the different DNA binding modes of DAPI with respectively, to the flow direction in a Couette cell with light respect to structure and to study how already bound DAPI propagating radially through the two concentric silica cylinders molecules may influence additional binding. Spectroscopic (Nordtn et al., 1992). The LD data are evaluated as the data (circular and linear dichroism, absorption, and fluores- reduced linear dichroism, LD', defined as cence) are presented for DAPI complexes, at various binding ratios, with [poly(dA-dT)]2 and with a few oligonucleotides LD'(X) = LD(X)/A,,(X) designed to address some specific questions regarding the where Ais,(X) is the absorption spectrum for the isotropic binding. sample in the absence of flow. This dimensionless quantity Our results are consistent with a model in which DAPI is related to the orientation of the light-absorbing transition binds cooperatively to the minor groove of [poly(dA-dT)] 2 moment as and in which the binding of several drug units leads to a ~(COS'a) - 1 conformational change of the polynucleotide. At high degrees LD'=SO=S(3 (3) of occupancy, crowding in the groove leads to the exciton type of interactions between adjacent DAPI molecules. The where S is an orientation factor describing the degree of spectral features of the DAPI complexes with the dodecamer orientationof the DNA helix in thegradient shear. It depends [d(CGCGAATTCGCG)]2 (also studied as a crystal), the on the stiffness of the DNA, the flow rate, and the viscosity alternating AT duplex, and natural DNA are very similar at of the medium (Nordtn et al., 1992). The optical factor, 0, low binding ratios, confirming similar minor groove binding reflects the angle, a, between the transition moment of in all of these cases. In addition, the binding of DAPI to absorption and the local DNA helix, and the brackets denote oligonucleotides containing AT sequences of varying lengths an ensemble average over the angular distribution. When indicates that sequences longer than four AT base pairs give several transitions are involved, the measured LD' will be an a cooperative behavior while a short run of four AT base pairs average of the LDTvalues of the individual transitions weighted behaves like a single site, in agreement with recently reported with their respective absorbances. The a-a* transitions of data on a decamer (Loontiens et al., 1991). the DNA bases, which are responsible for the LD of the polynucleotide around 260 nm, are polarized in the plane of MATERIALS AND METHODS the bases with an effective angle, a = 86O, between the planes of the bases and the helix axis (Matsuoka & Nordkn, 1982, Chemicals. [poly(dA4T)];!was obtained from Pharmacia, 1983). This angle is used to calculate the orientation factor, and the oligonucleotides were synthesized on an Applied S,from the measured LD'of the DNA bases. The LD spectra Biosystems 381A DNA synthesizer and purified by FPLC. were measured on a Jasco 5-500 spectropolarimeter as The purity of the oligonucleotides was in most cases checked described elsewhere (Nord6n & Seth, 1985). by electrophoresis. DAPI was purchased from Serva and was Circular dichroism is the differential absorption of circularly used without further purification. All concentrations were polarized light: determined spectrophotometrically using the following molar absorption coefficients: €262 = 6600 M-' cm-' for [poly(dA- CD(X) = A,(X)- A,(X) (4) dT)]2; €342 = 27 000 M-' cm-l for DAPI in water (Kapuscinski where A1 and A, are the absorbances measured with left and & Skozylas, 1978). For all synthetic oligomers, the same right circularly polarized light. DAPI itself is not chiral but value of €260 = 6600 M-' cm-' was used, which we estimate acquires CD when bound to the chiral DNA. The mechanism to give correct concentrations within 10%. All nucleic acid behind this induced CD is not clear in detail but is believed concentrations are expressed in nucleotide bases, and oligomer to be an effect of nondegenerate coupling between transitions sequences are presented as 5' - 3'. of DAPI and transitions of the bases of the nucleic acid host The mixing ratio, R,is defined as the total number of added (Kubista et al., 1989; Lyng et al., 1991). All CD spectra were DAPI molecules per nucleotide base (for the polynucleotides) measured on the Jasco 5-500 spectropolarimeter using a 1-cm alternatively per oligomer. It may deviate somewhat from quartz cell. the binding ratio, which is the number of bound DAPI Absorption spectra were recorded on a Varian Cary 2300 molecules per base. However, since the binding of DAPI to spectrophotometer and fluorescence spectra on an Aminco AT sequences is very strong at low ionic strength, R is SPF-500 spectrofluorometer in the quantum-corrected mode. essentially equal to the true binding ratio under our conditions All spectra were corrected for the inner filter effect. (10 mM Tris), even at relatively high mixing ratios. Reported Normalization. The titrations were made either at constant association constant in 100 mM NaCl is 2.9 X lo7 M-' with DAPI concentrations or at constant nucleotide concentrations, [poly(dAAT)]2 (Loontiens et al., 1991). As confirmed from but the presented data (except for LD) have been normalized the absorption spectra, the amount of free DAPI did not in with respect to the (total) DAPI concentration. any case exceed 20% and was in most cases far below this level. RESULTS Unless otherwise stated, the experiments were conducted in 10 mM Tris and 1 mM EDTA (pH 7.4). Measurements with the oligonucleotides were performed at 2-3 OC to LD at Low Binding Ratios. Figure la-c shows the linear minimize strand separation, and measurements with long DNA dichroism, as well as the absorption and the calculated reduced were made at room temperature. linear dichroism, at various DAPI/DNA mixing ratios. The Allosteric Binding of Ligands to DNA Biochemistry, Vol. 32, No. 12, 1993 2989
DAPI + [poly(dAdT)],
a) Rfpstwe 40. - 1 .I fA A 0. -0.0042 B 0.010 C 0.050 A D 0.10 - - 0.12 E 0.20 F 0.30
.., . . . . . 0.0042
--- 0.19 k - - 0.33 5 \ 1.0
a .oo d 5 - - 0.33 -.a
2w 480 Wavelength / nm FIGURE1: Absorption (A), linear dichroism (LD), and reducedlinear dichroism (LD‘) ofDAPI in the presenceof [poly(dA-dT)]~atvarious Wavelength I nm DAPI per DNA base mixing ratios (R). FIGURE2: Circular dichroism (eI - fro),absorption coefficient (e), and fluorescence of DAPI and [poly(dA-dT)]2 at various DAPI per LD is negative for the polynucleotide in the PT* transition DNA base mixing ratios (R). Data normalized to unit DAPI region around 260 nm, in accord with the planes of the bases concentration. in B-DNA being perpendicular to the helix axis. For DAPI the LD is positive in the 300-400-nm region, where DAPI is some free DAPI, which does not contribute to the LD but known to have two electronic transitions polarized near the contributes to the absorption in the calculation of LDT(cf. eq long axis of the molecule (Kubista et al., 1989). With the 2). A similar decrease of the LD signal also occurs at 270 orientation factor S calculated from the reduced dichroism nm and can be attributed to the presence of free DAPI, but of the DNA absorption band at 260 nm, the LD‘ of DAPI it is in part also due to absorption from bound DAPI. The gives an effective angle a of 43-46’, varying slightly from the orientation of DAPI relative to the DNA helix axis thus seems long to the short wavelength end of the absorption region to be essentially independent of the binding ratio. 300-400 nm. The contribution from DAPI at 260 nm is CD at Low and Intermediate Binding Ratios. While the negligible compared to the nucleic acid absorption at low LD spectra reveal no significant difference in the orientation binding ratios. The orientation of DAPI with the alternating of DAPI between low and intermediate binding ratios, the AT duplex is essentially the same as that concluded with calf CD spectra show marked variations (Figure 2a). At very low thymus DNA at a low binding ratio (Kubista et al., 1987). binding ratios, the DAPI CD spectruminduced by the presence According to a crystal study of DAPI bound to an oligonu- of the polynucleotide consists of a positive feature with a cleotide, the tile angle is 45O (Larsen et al., 1989), which maximum at 335 nm. An identical spectrum is observed when agrees very well with our results. DAPI binds to natural DNA at very low ratios, and it has LD at High Binding Ratios. As more and more DAPI been denoted as the “type I spectrum” (Manzini et al., 1983). binds to the polynucleotide, the orientation factor S increases However, at R = 0.02, corresponding to one DAPI per 50 as judged from the increase of ILDrl for DNA at 260 nm, as bases, a new and more intense induced CD band already well as for the DAPI band at h > 300 nm. The binding of appears with a maximum at 375 nm, providing evidence for DAPI at these higher ratios thus seems to lead to a stiffening the presence of a component denoted as the “type I1 spectrum”. and/or a lengtheningof the polynucleotide. A virtual change The spectral transition displays an isosbestic point at 341 nm. in shape of the LDr spectrum of DAPI at X > 300 nm at the It should be noted that no corresponding structural change is highest mixing ratios is mainly a result of the presence of evidenced by any other spectroscopic technique at these low 2990 Biochemistry, Vol, 32, No. 12, 1993 Eriksson et al. binding ratios. Clearly, CD is more sensitive to variations in long wavelengths, evidencing exciton interactions between binding geometry than the other spectroscopic properties and bound DAPI molecules. is able to detect structural changes associated with multiple Absorption and Fluorescence. The binding of DAPI to the binding. AT dodecamer is accompanied by a decrease of about 20%, CD at High Binding Ratios. As R increases to about 0.15, a red shift of some 20 nm of the absorption spectrum (Figure a new (negative) CD feature builds up at X > 400 nm. This 3b), and a 40-fold increase in fluorescence intensity (Figure feature has its counterpart in changes in absorption and 3c). These changes are comparable to those observed upon fluorescence (see below) and is characteristic for strong binding of DAPI to [poly(dAdT)]2,suggesting similar modes electronic interactions between the ligands. We interpret it of interaction. These features remain constant up to a mixing as an exciton type of interaction between DAPI molecules ratio of two DAPI molecules per oligomer, whereafter a forced to be bound close to each other due to crowding on the decrease and red shift of both absorption and fluorescence polynucleotide substrate. intensity are seen, evidencing exciton interactions between Absorption. Figure 2b shows the absorption spectra of free bound DAPI molecules. and bound DAPI. The binding of DAPI to the polynucleotide is accompanied by a hypochroism of about 12% and a red [d(ATATAT)]2 shift of the absorption maximum above 300 nm by about 20 CD. With the AT hexamer the CD titration turned out nm. These features remain constant up to a binding ratio of differently (Figure 3d). At the lowest binding ratio the CD 0.12, Le., within the range where large changes are observed spectrum is close to a type I spectrum, but there is no conversion in the CD spectra. At higher mixing ratios, however, the into a type I1 spectrum as the mixing ratio is increased. Instead, absorption decreases drastically and shifts to even longer there is a negative feature at 400 nm, reflecting exciton wavelengths as a result of the mentioned DAPI-DAPI interactions between DAPI molecules and showing that more interactions. than one DAPI molecule can bind to the hexamer. This Fluorescence. There is a large increase in the quantum oligomer is apparently too short to promote cooperative binding yield and a narrowing of the shape of the emission spectrum to AT mode 11, which evidently requires at least two bound when DAPI binds to [poly(dA-dT)]2 (Figure 2c). As with units possibly at a certain minimum distance from each other. the absorption, the fluorescence intensity is constant up to R Absorption. The absorption spectrum of the singly bound = 0.12. At higher ratios the emission decreases significantly DAPI molecules is not identically the same as that with the and shifts to longer wavelengths. Again this behavior is longer AT oligo- and polynucleotides (Figure 3e). The consistent with exciton interactions between closely bound hypochromism is similar, about 20%, but the red shift is less, DAPI molecules. about 10 nm, and the spectral shape is somewhat changed. From these results we conclude that DAPI binds to [poly- This could be an effect of the fact that the hexamer is too (dA-dT)]2 in two different modes, the “AT mode I” and the short to be completely in duplex form and/or that the end “AT mode 11”. The two binding modes are directly related effects are large. Indeed, a small decrease of the absorption, to the type I and type I1 CD spectra, respectively. For both by a few percent, is observed in the DNA absorption band at binding modes DAPI is bound with its long axis at about 45’ 260 nm when DAPI is added to the oligomer solution (results to the polynucleotide axis. In the AT mode I, DAPI molecules not shown). This can be understood as a shift of the are bound isolated, whereas the AT mode I1 requires the equilibrium between single- and double-stranded oligonucleo- presence of previously bound DAPI molecules, which results tides due to a stabilization of the duplex as DAPI binds. Such in a higher binding affinity and a stiffening of the DNA fiber. a stabilization is also evidenced from a nearly 40 OC elevation At very high mixing ratios we see evidence of Coloumbic of the melting temperature of [poly(dAdT)]2 upon binding (exciton) interactions between bound DAPI molecules. of DAPI (data not shown). Fluorescence. The fluorescence emission increases upon [ d(ATA TA TA TA TA T)] 2 binding just as with the dodecamer, but a decrease, and a CD. The magnitude of the induced CD of DAPI bound to slight red shift, with increasing R is observed for the entire the AT dodecamer is some 40% smaller than the CD of DAPI interval studied (Figure 3f). bound to [poly(dA-dT)]2, and the spectral shapeis somewhat [ d (CGCGAATTCG CG)] 2, [ d (CGCGATA TCGCG) ] 2, and different (Figure 3a). These differences can most likely be [d(GCATATGC)]2 ascribed to the short length of the oligomer and to fraying of the ends, since the strength of the induced CD depends on the CD. Up to a mixing ratio of one DAPI per oligomer, the length of the duplex substrate (Kubista et al., 1989; Lyng et shapes of the CD spectra of DAPI bound to the dodecamers al., 1991). The principal behavior of the oligomer is similar, [d(CGCGAATTCGCG)]2 (the Drew-Dickerson dodecamer) however, to that of the AT polymer: the CD of the bound (Figure 4a) and [d(CGCGATATCGCG)]z (Figure 4b), as DAPI changes gradually from a type I to a type I1 spectrum, well as to the octamer [d(CGATATGC)]* (Figure 4e), are evidencing initial binding in AT mode I and subsequent binding all very similar to that of DAPI bound to [poly(dAdT)]2 at in AT mode 11. The change of binding mode is already readily a low degree of occupancy. Only the strength of the CD is observed at low binding ratios, as expected for cooperative weaker by nearly a factor of 2. The constancy in shape binding. For example, at a mixing ratio of one DAPI per demonstrates that there is no cooperative binding to these oligomer, the recorded CD spectrum has a clear type I1 oligonucleotides. As R increases above one DAPI per character, meaning that a considerable amount of the DAPI oligomer, a negative dip at 400 nm appears, implying molecules are bound in AT mode 11. Since the AT mode I1 interaction between bound DAPI molecules. Whether the requires the presence of more than one DAPI molecule per increase in CD around 370 nm is due to exciton interaction oligomer, the bound DAPI molecules must be unevenly alone or whether mode I1 binding is also possible to these distributed among the oligomers, as is expected for cooperative oligomers, despite the short AT sequence, is not possible to binding. At even higher mixing ratios (more than two DAPI settle. Binding to the GC sequences may also be anticipated, molecules per oligomer) a negative CD feature appears at but the CD from such complexes should be relatively weak Allosteric Binding of Ligands to DNA Biochemistry, Vol. 32, No. 12, 1993 2991 DAPI + [d(CGCGAATTCGCG)], DAPI + (d(GCATATGC)], M
-0.33 ...... 1.0 --- 1.3 - - 2.5
- 0.33 ...... 1 .o
D
Wavelength 1 nm 9) R 1psr oligoma DAPI + [d(CGCGATATCGCG)], -0.33
4 --- 1.3 R/PPOli@OlDCI - - 2.5 ...... 0.96
400 Wavelength 1nm
...... 0.96
- - 3.6
Wavelength nm
FIGURE 3: Circular dichroism (CI - er), molar absorptivity (e), and fluorescence of DAPI and [d(CGCGAATTCGCG)], (a,, b), [d(CGCGATATCGCG)]2 (c, d), and [d(GCATATGC)]z (e-g) at various DAPI per oligomer duplex mixing ratios (R). Data normalized to unit DAPI concentration. as judged from the CD of DAPI bound to [poly(dG-dC)]z one DAPI bound per oligomer molecule the absorption is (Nordbn et al., 1990). According to the CD behavior, as further decreased and shifted toward longer wavelength as a many as 3-4 DAPI molecules can bind to each dodecamer, result of exciton interactions between adjacent DAPI mol- and at least two DAPI molecules can bind to the octamer. ecules. Absorption. The shapes of the absorption spectra of DAPI Fluorescence. DAPI bound to the octamer displays a 30- when bound to the oligomers at low ratios are similar to that fold increase in fluorescenceemission intensity and a narrowing of the complex with [poly(dAAT)]~,but with absorptivities of the spectrum (Figure 4g). As R approaches one DAPI per that are somewhat smaller (Figure 3b,d,f). With more than oligomer, the intensity starts to decrease and a new emission 2992 Biochemistry, Vol. 32, No. 12, 1993 Eriksson et al.
DAPI + [d(ATATATATATAT)], DAPI + [d(ATATAT)], so. 4 RI~zc~wnw - 0.25 20. - --- 0.75 -- 2.0
-'Osw' ' ' ' ' ' ' ' ' ' ' ' ' ' sw 403 480-
4 R PS ohgomu R 1 pr otipma a - 0.48 - 0.25 .4 \ ...... 0.96 .-,x 24 - - 2.0 2 --- Y - - 4.8 .-
I8 -G.
4w OW 403 OW Wavelength nm Wavelength 1 nm
FIGURE4: Circular dichroism (el- tr),molar absorptivity (e), and fluorescenceof DAPI and [d(ATATATATATAT)I2(a-c) and [d(ATATAT)]* (d-f) at various DAPI per oligomer duplex mixing ratios (R). Data normalized to unit DAPI concentration. arises at lower wavelengths, again demonstrating interactions is not due to any specific sequence of DNA. We shall argue between DAPI chromophores that are bound on the same that the second site is a conformational effect due to an oligonucleotide. allosteric binding of DAPI in the minor groove. This hypothesis (Nord6n et al., 1990) has support in binding DISCUSSION isotherm studies which are consistent with a strongly coop- erative binding of DAPI to [poly(dA-dT)]* (Wilson et al., In this work, we address a number of questions regarding 1990). the details of the binding of the dye 4',6-diamidino-2- phenylindole to DNA. It is clear from our results that the The flow linear dichroism results (Figure la+) show that binding mode depends on the binding stoichiometry as well DAPI bound to [poly(dA-dT)]z is oriented with its long axis as on the DNA sequence, in agreement with the findings by polarized transitions at an angle of 43-46' relative to the Wilson et al. (1990). Our findings suggest, as will bediscussed helix axis. The constant LD' over the 320-400-nm region is below, how a ligand, when bound contiguously to DNA, can consistent with a near-planar DAPI conformation in the induce alterations of the DNA structure. As a model system complex. This angular orientation is in good agreement with DAPI may provide insight important for the understanding the crystal structure of the DAPI complex with [d(GCG- of potential long-range actions in DNA-drug and DNA- CAATTCGCG)] 2, where DAPI is sitting edgewise inserted protein complexes in vivo. into the minor groove at 45O to the helix axis (Larsen et al., BindingofDAPItoATRegions. Aproblemin earlystudies 1990). of the binding of DAPI to DNA has been the interpretation That the drug is residing deep in the groove is evidenced of the "second binding site", evidenced from changes in the by the observation of a complete protection against quenching circular dichroism spectrum at binding ratios around 0.02 in of the fluorescence by iodine (S. K. Kim et al., unpublished DNA. A crucial finding is the observation that a homopoly- results) as well as by iodide ion (NordBn et al., 1990). nucleotide, [poly(dA-dT)]z, also shows the same behavior, Additional evidence that the binding geometry of DAPI in demonstrating that the virtual low abundancy of the first site [poly(dA-dT)]2 is very similar to that found in the crystal is Allosteric Binding of Ligands to DNA Biochemistry, Vol. 32, No. 12, 1993 2993 our observation of almost identical induced circular dichroism spectra of DAPI when bound to this polynucleotide and to the oligomer used in the crystal study. We also note that the same CD spectra are observed when DAPI is bound to the t dNg #. dNg oligonucleotides [d(CGCGATATCGCG)]2 and [d(GCATAT- I I I I I GC)]2 at stoichiometries below one drug molecule per oligonucleotide. In other words, the same type of minor groove I ImI I I binding geometry is found whether DAPI is bound to an ATAT or to an AATT sequence. + drug #.drug Wih increasing binding ratios, neither of the AT-type DAPI complexes exhibits any major change in binding geometry. In [poly(dA-dT)] 2 DAPI is, according to linear dichroism, bound at an angle of about 45O, consistent with minor groove binding and independent of binding ratio. Also the absorption and emission features indicate that there are no significant variations in the binding mode. By way of contrast, strong variations in circular dichroism clearly demonstrate that some changes occur. The fact that the CD changes are observed already at rather low mixing ratios gives an indication that +drug #-drug they may not be due to interactions between bound species that are accidentally close to each other but that this is more likely to be a cooperative phenomenon. A convex binding isotherm has been reported by Wilson et al. (1990) for DAPI with [poly(dA4T)12,consistent with cooperative or allosteric + drug 4f. drug binding. Furthermore, the appearanceof a type I1 CD, which is not accompanied by corresponding changes in the absorption and fluorescence spectra, may indicate that the type I1 is not due to the different molecular conformation itself. The different conformations of DAPI would interact with DNA in different ways; hence, the changes in the absorption and fluorescence spectrum would be expected. We may explain the notable CD behavior in such terms as follows. clllll Binding Model for Cooperative Allosteric Binding of DAPI to AT Regions. Figure 5 schematically shows three possible t drug if.drug simplistic models for cooperative binding. Model A, which does not involve any structural change of the DNA lattice, assumes that the extra binding affinity for a second DAPI molecule stems from attractive interactions between the two bound ligands. However, the DAPI molecules carry a divalent + drug #.drug positive charge and should repel1 each other electrostatically. Further, if the DAPI molecules had been closely interacting with each other, their electronic structures should have been perturbed, and this would have been revealed by changes in FIGURE5: Three models for cooperative binding (see text for the absorption and fluorescence spectra. Indeed, spectral explanation). changes evidencing Coulombic (exciton) interactions between bound ligands are observed, however, first at very high binding agreement with our results. How extensive this allosteric ratios, far beyond those for the AT mode I1 binding. We change of the DNA structure is cannot be deduced solely therefore discard scheme A as important for any cooperative from our results. No significant change of the orientation of binding of DAPI to [poly(dAdT)]z. DAPI is seen, indicating that the helicity and pitch of the In model B, the binding of the first DAPI molecule is duplex are effectively retained. On the other hand, the large assumed to induce some conformational change of the DNA increase of the LDr of the polymer shows that the binding lattice, with the result that the additional DAPI will bind with leads to a considerable stiffening of the structure,which might somewhat greater affinity. In this scheme all DAPI molecules, be related to such an allosteric conformational change, but both those bound isolated and those bound cintiguously, sit which also might be an effect of DAPI filling the minor groove on the same, perturbed lattice. This scheme can also be and decreasing the bendability of DNA. discarded, since isolated and contiguous DAPI molecules are The CD spectra of DAPI bound to [poly(dA4T)]2 (Figure bound differently as evidenced from their distinctly different 6a) were analyzed using a chemometric approach (see CD spectra. Appendix l), allowing an estimate of the pure spectrum of the In model C the isolated DAPI molecules bind to an induced CD of DAPI in the AT mode I1 and the amount of unperturbed lattice, but when bound contiguously they induce DAPI bound in the two modes at different binding ratios. The some conformational change of the lattice which is favorable type I1 CD spectrum can be deduced only within certain limits for binding and gives a greater affinity. Note that in this (see Appendix 1 for an explanation), and in Figure 6 two scheme, the isolated DAPI molecule is bound in the AT mode cases are depicted. A criteron for Figure 6d,e was that the I; however, when additional DAPI molecules bind contigu- analysis should produce a mode I1 spectrum that has reasonable ously, the ligands become bound in the AT mode 11, in shape and intensity. As seen in Figure 6e, the amount of 2994 Biochemistry, Vol. 32, No. 12, 1993 et al.
4
eo. -
0 Wavelength nm
Lo . ._ C --. I- b, I
;
6 Wavelength / nm
.15
Total binding &io FIGURE6: CD spectra of DAPI bound to [poly(dA-dT)]2after noise reduction by chemometric treatment (a). CD spectra of DAPI bound in mode I (-) and mode I1 (- - -) (b). Corresponding binding isotherms for mode I (*) and mode I1 (0)calculated by chemometry compared to binding isotherms for random occurrence of isolated DAPI molecules (-) and DAPI molecules with at least one neighbor (- - -) (c). The R,, parameter, and thereby the mode I1 spectrum, was tuned to give best fit of CD data to the case of random binding. Another possible solution by the chemometric approach where RI~is tuned to give an all-positive mode I1 CD spectrum with a physically reasonable shape (d, e). Here, the calculated binding isotherms predict a larger occurrence of contiguously bound DAPI ligands than expected from random association.
DAPI in AT mode I initially increases linearly with increasing Binding of DAPI to Oligonucleotides. Let us see how this total amount of bound DAPI and reaches a maximum at about model applies to the interpretation of binding of DAPI to the 0.09 DAPI molecule per DNA base. The amount of DAPI oligonucleotides. For the AT dodecamer, just as with [poly- in the AT mode I1 is initially very low but increases roughly (dA-dT)]2, increasing binding ratio leads to a type I1 CD quadratically with the amount of DAPI added. The binding spectrum with a positive maximum around 375 nm. The isotherms indicate that molecules initially bound in the AT change in CD is observed at a binding ratio of much less than mode I are converted into the AT mode I1 as the titration one drug per oligonucleotide, which is consistent with a proceeds, as predicted by scheme C. (The thermodynamics cooperative binding. The dodecamers [d(CGCGAAT- of the allosteric binding is discussed in more detail in Appendix TCGCG)]2 and [d(CGCGATATCGCG)]2 both show, at low 2.) Allowing that a mode I1 peak is somewhat larger, but binding ratios, a CD feature characteristic of mode I, and at definitely not unphysical, gives a perfect match between the higher binding ratios, there is evidence for a second mode of two modes of binding and the occurrence of isolated ligands binding. However, in these cases no cooperativity is observed, and ligands with at least one neighbor, as calculated by the but the "switching point" occurs around one drug molecule model of McGhee and von Hippel (1974) assuming no bound per oligonucleotide unit. The same holds for the cooperativity (Figure 6b,c). The chemometric analysis cannot octamer [d(GCATATGC)] 2. unambiguously point out a binding mechanism but can at The different behavior in the shorter AT tracts can beviewed least hint at a possible solution. We see that cooperative as an effect of two DAPI molecules being forced to bind more binding and accidental neighbors are both possible. closely together, thereby losing some of the free energy of Allosteric Binding of Ligands to DNA Biochemistry, Vol. 32, No. 12, 1993 2995 binding due to mutual electrostatic repulsion by their divalent the DNA conformation is changed to favor binding. The new cationic charge. Indeed, the appearance of an exciton type conformation is responsible for the appearance of the type I1 of interaction of circular dichroism at low binding ratios, as induced CD spectrum of the bound DAPI chromophore. well as corresponding changes in absorption and fluorescence, Chemometric analysis of the CD spectra of solutions with shows that the two DAPI molecules are closer than in the varying relative contents of the different spectroscopic species longer AT tract complexes at a corresponding binding ratio. supports this model and suggests that the DNA conformation This observation provides an important corollary as to the switches when at least two or three DAPI ligands are bound minimum size of the drug-drug distance in the normal contiguously. “dimeric” type I1 binding mode where no such interactions 5. Results from the oligonucleotide complexes studied and are observed. evidence about the proximity of adjacent DAPI molecules at Exciton Inreraction. In the absorption, linear and circular high binding ratios from exciton interactions suggest that the dichroism, and fluorescence spectra of the binding of DAPI DAPI interactions, which are found in longer AT sequences, to [poly(dA-dT)] 2 and the AT-containing oligonucleotides, are of long-range nature, potentially with several base pairs there is evidence for electronic (exciton) interactions between separating the drug units. adjacent DAPI molecules at high binding ratios. The LD spectrum of the DAPI-[poly(dA-dT)] 2 complex shows a APPENDIX 1 positive feature at 380 nm followed by a negative trend at the long-wavelength edge of the absorption (the corresponding The CD spectra of DAPI bound to [poly(dA-dT)]z were absorption spectrum shows a tail evidencing the exciton analyzed by a chemometric approach (Kubista, 1990). The interaction). The LD spectrum in principle contains infor- CD spectra of the DAPI absorbing region at 300-450 nm and mation about the structure of this “dimer”. For example, mixing ratios 50.12, for which the exciton contribution to the with two DAPI molecules lying in the minor groove in a head- spectra is still negligible, were normalized to 1 M total DAPI to-tail arrangement, the low-energy component of the exciton concentration and arranged as rows in an n X m matrix A, couplet would correspond to a transition moment oriented at where n = 12 is the number of CD spectra used in the analysis about 30’ to the helix axis. This should give an LD that is and m = 751 is the number of data points for each spectrum. more positive than the 45’ LD of a single minor groove bound Since the spectral response of each component is proportional DAPI molecule. The high-energy component (to appear at to its concentration (Beer-Lambert law), matrix A must equal shorter wavelength) would give a weak negative LD in this the product CV’: case. Another possibility, a stacked typeof DAPI dimer (such as the dimers inferred for chromomycin (Gao & Patel, 1989) A = CV’ (AI) and distamycin A (Pelton & Wemmar, 1990)),would instead where V’ is an r X m matrix of the molar CD spectra of the give a transition moment for the low-energy component which r pure components, and C is an n X r matrix containing the is perpendicular to the helix axis and would thus be associated concentrations of these components in the n samples. with a negative LD. The high-energy component would give The data matrix A is decomposed into principal components a strong positive LD for this case. The experimental spectrum using the NIPALS algorithm (Fisher & MacKenzie, 1923): is clearly more consistent with the latter geometry, which should then imply that quite a high number of DAPI molecules A = TP‘ (‘42) could be inserted in the minor groove, a conclusion supported where P’ contains the projection vectors, Le., the principal by the very high maximum binding ratios observed for this components, as rows and T contains the target vectors as drug in [poly(dA-dT)] 2. A recent simulation of such a binding columns. The product TP’ will reproduce A, but the P’ matrix geometry of DAPI in the minor groove shows that such a will not be equal to the V’ matrix; its rows are linear geometry is indeed possible (Mohan & Mathindra, 1991). combinations of the true spectra of the present components. Likewise, T will not equal C but tells only how the principal CONCLUSIONS components should be combined to reproduce the measured spectra of A. Inspection of P’ and T revealed that two The present optical spectroscopicstudies of the complexation components are sufficient to account for all of the spectral of DAPI with polynucleotides and oligonucleotides have features in thedata matrix A. Thus the number of independent allowed a number of important conclusions to be drawn. spectral components in the samples is 2 (r = 2). By excluding 1. In AT-rich sequences, to which DAPI binds preferen- all but the first two rows in P’ and the first two columns in tially, the molecule is edgewise inserted into the minor groove T,we eliminate most of the noise in the experimental data and with its long axis at an angle of approximately 45” to the helix retain only the physically relevant features (Figure 6a). axis. This binding geometry is found for very low as well as Equation A2 is one possible decomposition of A into a higher binding ratios. product of two matrices, but so also is eq A3, where R is any 2. The concluded geometry is in agreement with the arbitrary nonsingular r X r matrix: geometry of the DAPI complex observed in the crystalline form with the Drew-Dickerson dodecamer, and the latter A = (TR)(R-’P’) (A31 complex is verified to give an induced CD spectrum very similar The problem is now to find a matrix R such that to that of the complex with [poly(dA-dT)Jz in solution. 3. Our observation of two distinct types of CD spectra, TR=C (A41 typical for the binding of DAPI to DNA and for the complex and with [poly(dA+lT)]z, demonstrates that the first binding mode, despite its very low apparent abundance, is not due to a specific R-’P’ = V 645) DNA site. For our system it is not possible to solve these equations 4. The two binding modes could be explained in terms of unambiguously. However, with certain assumptions and an allosteric binding of DAPI to [poly(dA+lT)]z such that, judgement and reasonable spectra of the pure component a when DAPI molecules bind contiguously to the AT sequence, fair estimation can be made. Since we use only the first two 2996 Biochemistry, Vol. 32, No. 12, 1993 Eriksson et al. principal components, R is a 2 X 2 matrix: out on the basis of the chemometric analysis. Figure 6d shows a different solution where R11 was tuned to give a type I1 CD R = [::: and R-' = (RllR2,- R12R21)-1[:& i::z] spectrum that is all positive and has a more traditional shape. The corresponding binding isotherms predict a larger abun- (A61 dance of DAPI ligands bound in mode I1 than expected from All CD spectra used in the analysis cross in an isodichroic random occupancy, which is in agreement with an allosteric point, which means that DAPI is quantitatively bound in these binding process that leads to a clustering of bound ligands. experiments. Since the spectra were scaled to unit DAPI From the ahalysis we conclude that, even though our data concentration, the sum of the concentration coefficients for cannot prove allosteric binding of DAPI to [poly(dA4T)]2, each sample must be equal to 1: they are clearly consistent with such a mechanism. Cli+ C2i= 1 (i = 1,n) (A7) Equations AkA7 then give APPENDIX 2
Thermodynamics of Allosteric Binding. The CD titration or data are consistent with the binding model C in Figure 5, where an isolated DAPI molecule does not induce a confor- mational change of the DNA helix, but contiguously bound where Til and Tj2 are the ith elements in columns one and two, DAPI molecules do. Such a binding scenario can be respectively, in matrix T. By fitting the two columns of T to rationalized by an allosteric binding model. a vector where all elements are 1, we determine F11 = (R11 + R12) and F12 = (R21 + R22). The thermodynamics for allosteric binding to DNA was The CD spectra at very low binding ratios are, within originally described by Dattagupta et al. (1 980). The DNA experimental error, independent of the amount of bound DAPI, latticeunits, the base pairs, are assumed to be in an equilibrium and we assume that these spectra represent the spectrum of between two states to which the ligand binds with different DAPI bound in the AT mode I. By letting the CD spectrum affinities. These states are the standard B-form DNA of the sample with the lowest binding ratio be the first row (represented by white boxes) and a different state of higher of matrix A, we have V'li = Alj (i = 1, m) and energy (represented by shadowed boxes). The free DNA lattice units are then in equilibrium between these two states. ~22(~11~22- ~12~21)-'~'1i -
~12(~11~22- ~12~21)-'~5j = A'lj (~8) or where P'lj and Pij are the ith elements in rows one and two, The thermodynamicparameters describing this equilibrium respectively,in matrix P'. By fitting the two projection vectors obtained from the NIPALS algorithm to the CD spectrum and ligand binding to the two states are the following: recorded at the lowest binding ratio, we determine F21 = R22(RllR22 - R12R21)-l and F22 = R12(RllR22 - R12R21)-l. S Out of these four parameters F11, F12, F21, and F22, only three are linearly independent, and we cannot determine all four elements in the R matrix. However, we can use these relations to express matrix R with a single parameter:
This means that for any given value of R11 that does not make matrix R singular, we can calculate a mathematically acceptable solution for C and V, where the AT mode I CD spectrum is recorded at the lowest mixing ratio and where the total concentration of DAPI in all samples sums to 1. However, not all values of R11 provide physically acceptable solutions, n K2 since most values of R11 predict negative concentrations in some of the samples. By requiring that the concentrations of the two components be nonnegative in all samples, we can considerably limit the number of solutions. Figure 6b,c shows where AG," = -RT In s is the cost in free energy when one solution consistent with our data. Here R11 was tuned to transforming 1 mol of base pairs from the B form to the high- give binding isotherms similar to those of fandom binding, where isolated bound DAPI forms one class of ligand and energy state, AGj" = -RT In t~ is the cost in free energy when contiguously bound DAPI molecules form a second class. The creating 1 mol of junctions between lattice segments of the CD spectrum of the second form, consistentwith these binding two states, and K1 and K2 are the binding constants of the isotherms (Figure 6b), has a negative feature at short ligand to the two states. In addition, a general treatment wavelength which may be hard to rationalize. The spectrum should include cooperativity interactions between the bound is however not unphysical, and this solution cannot be ruled ligands: Allosteric Binding of Ligands to DNA Biochemistry, Vol. 32, No. 12, 1993 2997
011 of the true length but rather is an indication of an increase in the persistence length of the DNA. "mFmPlm The structural change is related to the minor groove binding mode of DAPI and is not observed for the [poly(dG4C)]2 022 binding mode (NordCn et al., 1990). We note that the other ligands, netropsin and distamycin A (Hogan et al., 1979; Dattagupta et al., 1980), which induce similar conformational changes are also minor groove binders with a preference for 012 AT regions (Zimmer & Wahnert, 1986), and we shall henceforth refer to this altered DNA form as the m form. Since these minor groove binders interact much more strong111 with the m form of DNA (DAPI -90 times (Wilson et al., 1990);distamycin 15 times (Dattagupta et al., 1980))despite where AG,,,' = -RT In 00 is the change in free energy when - two bound ligands, one in state i and one in statej, are brought having a similar binding geometry as when bound to standard next to each other. B-DNA (Figures 1 and 2), the minor groove of the m form could be different. In the absence of ligands, the equilibrium between the two states is shifted far to the normal B form. However, if K2 > Long-Range Effects Mediated through DNA Conforma- K1, the ligands bind preferentially to the high-energy state of tional Changes. The cost in having a segment of DNA in the the lattice and may induce a conformational change. Our m form is mainly due to the creation of junctions and not due. results demonstrate that isolated bound DAPI molecules do as much to the energy loss of having the base pairs in the not induce the conformational change. Assuming a binding altered state. Therefore, if two segments in the m state are site of four base pairs (Portugal & Waring, 1988; Jeppesen separated by n free bases, the system may gain in energy if & Nielsen, 1989),the equilibrium constant between an isolated the intervening free base pairs are also converted into the m ligand bound to B-form DNA and to a local segment of the state, since the energy loss in having them in the altered state high-energy state is Kl/K2s4a2 > 1. We also find that is compensated for by the elimination of two junctions. The contiguously bound DAPI molecules do induce the confor- partition constant between all of the intervening base pairs in mational change in an AT dodecamer. This oligomer should the standard B form and all in the anomalous m form is sP/a2, have the capacity to bind three ligands. Since no junctions where p is the number of free intervening base pairs. Note are created, Kl3/Kz3sI2< 1. that the ratio only contains parameters describing properties Wilson et al. (1990) studied the binding of DAPI to [poly- of the DNA lattice and not properties of the ligands inducing (dA-dT)]2 by equilibrium dialysis and analyzed the data the m state. Dattagupta et al. (1980) obtained s = 0.962- according to the allosteric binding model, obtaining K2/K1= 0.977 (depending on ionic strength) and Q = 0.003 for the 90, s = 0.9 1, and Q = 0.002. Although their analysis neglected conformational change induced by distamycin A. The values cooperative interactions between the bound ligands, these of the thermodynamic parameters are similar to those for the values sbould provide rough estimates of the thermodynamic DAPI-induced conformational change (Wilson et al. (1990); parameters, and we calculate K1/K2#a2 = 4050 and K13/ see above), suggesting that the same conformational change K23s12= 4 X 10". Although the parameters were determined is induced by the two drugs and that these parameters indeed at a higher ionic strength than used in our study, the calculated are properties of the DNA lattice only. Using these values, ratios are consistent with the inequalities deduced from our we find that the equilibrium constant is unity for a stretch of observations. 132-500 free base pairs (depending on which values we use) Using the thermodynamic parameters, we can calculate enclosed by segments of m-DNA. This means that two how many DAPI molecules should be bound contiguously to m-DNA segments may efficiently convert some 13-37 turns induce the conformational change. The partition constant of intervening free DNA to the m form. between m contiguously bound DAPI molecules in the two DAPI binds only 90 times stronger to m-DNA than to states is For m = 3 this ratio equals 1.06, standard B-DNA (K2/K1= 90), and as discussed above, three meaning that a lattice segment covered by three contiguously contiguously bound DAPI ligands are expected to convert bound DAPI molecules is equally probable in either state. B-DNA into m-DNA. However, DAPI is a small ligand, and (Actually, the equilibrium will be shifted more to the altered one can imagine considerably larger K2/K1 ratio:. for other state by configurations where some neighboring free bases ligands. In order for a single ligand to convert DNA locally are also in the altered state.) into the m form, K2/K1 > s-*c2,where n is the binding size. The High-Energy State of DNA. Simultaneous with the With 5 < n < 15, the ratio K2/K1 must be on the order of lo5 appearance of a type I1 CD spectrum in the [poly(dA-dT)]2 to bring the lattice into the m form. However, for long-range titration, evidencing the structural conversion of the DNA signal transmission the equilibrium constant is not the sole lattice, we observe an increase in the overall LD magnitude. determinant; the rate of conversion is also important. In a This increase reflects a more efficient orientation of the DNA, possible real system a ligand may bind its binding site in which could be due to either a lengthening and/or a stiffening standard B-DNA, but not exert its action until the DNA is of the lattice (NordCn et al., 1992). The increase in orientation converted into the m form. An effector ligand bound in its is substantial and is unlikely to be entirely a lengtheningeffect, proximity may convert the DNA locally to the m form, and and we suggest that the high-energy state is considerably stiffer the structural change may transiently propagate to the first than normal B-form DNA. This is in accordance with the ligand and trigger its action. The rate of this process will structural change induce by netropsin (Hogan et al., 1979) depend on the rate and processivity of the structural conversion and distamycin A (Dattagupta et al., 1980),where the increase of the interveningsequence. In our experimentswe have never in the apparent length of the DNA was estimated to be 14 A observed signal changes in time. In all cases where a type I1 per bound ligand, which is clearly unrealistic as an increase CD spectrum was observed, it was seen immediately after 2998 Biochemistry, Vol. 32, No. 12, 1993 Eriksson et al. sample preparation, which means that the B - m transition Larsen, T. A., Goodsell, D. S., Cascio, D., Grzeskowiak, K., & cannot be a very slow process. Dickerson, R. E. (1989) J. Biomol. Struct. Dyn. 7,477-491. Whether this form of signal transmission over a distance Lee, G. M., Thornthwaite, J. T., & Rasch, E. M. (1984) Anal. occurs in real systems, we do not know. It has not been studied Biochem. 137, 221-225. very much, since there has been no way to detect this DNA Lin, M. S., Comings, D. E., & Alfi, 0. S. (1977) Chromosoma structural change. However, in DAPI we have a sensitive 60, 15-25. Loontiens, F. G., McLaughlin, L. W., Diekmann, S., & Glegg, probe for this structural change, since through its CD spectrum R. M. (1991) Biochemistry 30, 182-189. it will report the presence of m form induced by any other Lyng, R., Rodger, A., & Nordbn, B. (1991) Biopolymers 31, ligand. 1709-1720. Manzini, G., Barcellona, M. L., Avitabile, M., & Quadrifoglio, REFERENCES F. (1983) Nucleic Acids Res. 11, 8861-8876. Albinsson, B., Eriksson, S., Lyng, R., & Kubista, M. (1991) Manzini, G., Xodo, L., Barcellona, M. L., & Quadrifoglio, F. Chem. Phys. 151, 149-157. (1985) Nucleic Acids Res. 13, 8955-8966. Barcellona, M. L., & Gratton, E. (1990) Eur. Biophys. J. 17, Matsuoka, Y., & NordCn, B. (1982) Biopolymers 21, 2433- 3 15-323. 2452. Barcellona, M. L., & Gratton, E. (1991) Biophys. Chem. 40, Matsuoka, Y., & NordCn, B. (1983) Biopolymers 22, 1731- 223-229. 1746. Bella, J. L., & Gosklvez, (1991) Biotech. Histochem. 1,44-52. McGhee, J. D., & von Hippel, P. H. (1974) J. Mol. Biol. 86, Bernheim, A., & Miglierina, R. (1989) Hum. Genet. 83, 189- 469-489. 193. Mohan, S., & Yathindra, N. (1992) J. Biomol. Struct. Dyn. 9, Bierzynski, A., Boguta, G., Berens, K., & Wierzchowski, K. L. 695-704. (1978) Studia Biophys. 67, 57-58. NordCn, B., & Seth, S. (1985) Appl. Spectrosc. 39, 647-655. Brown, R. N., & Hitchcook, P. F. (1989) Dev. Brain Res. 50, Nordbn, B., Eriksson, S., Kim, S. K., Kubista, M., Lyng, R., & 123-128. Akerman, B. (1990) in The Jerusalem Symposia on quantum Cavatorta, P., Masotti, L., & Szabo,A. G. 91985) Biophys. Chem. Chemistry and Biochemistry (Pullman, B., & Jortner, J., Eds.) 22, 11-16. Vol. 23, pp 23-41, Kluwer Academic Publishers, Dordrecht, Chi, H., Ishibashi, Y.,Shima, A., Mihara, I., & Otsuka, F. (1 990) The Netherlands. J. Invest. Dermatol. 95, 154-157. Nordtn, B., Kubista, M., & Kurucsev, T. (1992) Q.Rev. Biophys. Dann, O., Bergen, G., Demant, E., & Volz, G. (1971) Liebigs 25 (1). 51-170. Ann. Chem. 749, 68-89. Pelton, J.' G., & Wemmer, D. E. (1990) J. Am. Chem. Soc. 11 2, Dattagupta, N., Hogan, M., & Chrothers, D. M. (1980) 1393-1399. Biochemistry 19, 5998-6005. Portugal, J., & Waring, M. J. (1988) Biochim. Biophys. Acta Fisher,R., & MacKenzie, W. (1923) J. Agric. Sci. 13,311-320. 949, 158-168. Gao, X., & Patel, D. J. (1989) Biochemistry 28, 751-762. Russel, W. C., Newman, C., & Williamsson, D. H. (1975) Nature Hajduk, S. L. (1976) Science 191, 858-859. 253, 461-462. Hard, T., Fan, P., & Kearns, D. R. (1990) Photochem. Photobiol. Schwartz, D. C., & Koval, M. (1989) Nature 338, 520-521. 51, 77-86. Szabo, A. G., Krajcarski, D. T., Cavatorta, P., Masotti, L., & Hogan, M., Dattagupta, N., & Crothers, D. M. (1979) Nature Barcellona, M. L. (1986) Photochem. Photobiol. 44, 143- 278, 521-524. 150. Jeppesen, C., & Nielsen, P. E. (1989) Eur. J. Biochem. 182, Takata, K., & Hirano, H. (1990) Acta Histochem. Cytochem. 43 7-444. 23, 679-683. Kapuscinski, J., & Skozylas, B. (1978) Nucleic Acids Res. 5, Tijssen, J. P. F., Beekes, H. W., & van Stevenick, J. (1982) 3775-3799. Biochim. Biophvs. Acta 721, 394-398. Kapuscinski, J., & Szer, W. (1979) Nucleic Acids Res. 6, 3519- Williamson, D. H.; & Fennell, D. J. (1975) Methods Cell Biol. 3534. 12. 335-351. Kapuscinski, J., & Yanagi, K. (1979) Nucleic Acids Res. 6,3535- Williamson, D. H., & Fennell, D. J. (1979) Methods Enzymol. 3542. 56, 728-733. Kubista, M. (1990) Chemom. Intell. Lab. Syst. 7, 273-279. Wilson, W. D., Tanious, F. A., Barton, H. J., Strekowski, L., & Kubista, M., Akerman, B., & Nordtn, B. (1987) Biochemistry Boykin, D. W. (1989) J. Am. Chem. SOC.111, 5008-5010. 26, 4545-4553. Wilson, W. D., Tanious, F. A., Barton, H. J., Jones, R. L., Fox, Kubista, M., Akerman, B., & NordCn, B. (1988) J. Phys. Chem. K., Wydra, R. L., & Strekowski, L. (1990) Biochemistry 29, 92, 2352-2356. 8452-8461. Kubista, M., Akerman, B., & Albinsson, B. (1989) J. Am. Chem. Zimmer, C., & Wahnert, U. (1986) Prog. Biophys. Mol. Biol. SOC.111, 7031-7035. 47, 31-1 12.